Back to models
LegacyChat Models

MiniMax

MiniMax M2-her

Best Overall

Legacy dialogue and role-play model with published 64K context and dedicated conversation settings.

DialogueRole-play
Open AppBrowse All Models

At a glance

Know the model before you prompt.

MiniMax API specifications
Context window
64K tokens (published)
Maximum output
2,048 tokens
Inputs → output
Text → Text
Knowledge cutoff
Not verified

API model ID: M2-her

Capabilities & boundaries

What it supports. Where the limits are.

Tool support requires the appropriate API integration; a supported tool is not automatically active in every chat.

Supported API features and tools

  • Text dialogue and multi-turn role-play in the dedicated M2-her guide
  • Provider-specific role settings including system, user_system and group context
  • Conversation examples and history to help maintain character and interaction style
  • Streaming dialogue through the dedicated documented chat request schema

Before you choose

  • The invocation guide publishes a rounded 64K context. The retained dedicated schema caps max_completion_tokens at 2,048; neither value establishes a knowledge cutoff.
  • Provider documentation is inconsistent: the dedicated guide names M2-her, while the general current Chat Completions schema lists other M-series IDs. Verify the correct endpoint and current access before integration.
  • The dedicated schema does not document tool calls, thinking controls or strict structured-output fields. Do not borrow these capabilities, or multimodal support, from M2.x or M3.
  • A convincing character response is not evidence of factual reliability, professional qualification or personal memory beyond the supplied conversation. Set boundaries and review sensitive interactions.

Reasoning behavior

The dedicated M2-her schema does not establish thinking switches or effort levels. Describe the desired dialogue and constraints directly; do not claim M2.x always-on reasoning or M3 adaptive controls for this separate model.

  • The retained dedicated schema uses model M2-her at POST /v1/text/chatcompletion_v2 on api.minimax.io. The linked general API documentation has changed, so validate this legacy endpoint against your account before integrating.
  • Do not assume Anthropic-compatible access: M2-her is absent from the reviewed supported-model list for that format. Follow the dedicated role/message schema instead of mixing provider message types.
  • Keep character setup, user role and group context separate from untrusted conversation content. Maintain explicit boundaries even if the persona prompt asks for a different tone.

Put it to work

Start with a more useful prompt.

Original examples from EZ Ai Assist. Adapt these to your task and the features available in your workspace.

Workflow 01

Define a bounded fictional character

Create consistent dialogue without hidden capability claims.

Create a dialogue specification for a fictional museum guide using this setting. Define voice, knowledge boundaries, how to handle unknown facts, and how to acknowledge that it is an AI character if asked. Include three sample exchanges and a rule for redirecting requests outside the museum topic. Do not invent real credentials or claim access to visitor records.

Workflow 02

Check continuity across a conversation

Find contradictions without manufacturing memories.

Review this conversation for character consistency. List facts established by the dialogue, later contradictions and unanswered questions. Cite the turns that support each finding. Propose a short next reply that acknowledges uncertainty naturally, preserves the agreed persona and does not pretend to remember anything outside the transcript.

Workflow 03

Rehearse a multi-person discussion

Keep roles and objectives distinct in a simulated exchange.

Simulate a short planning discussion between the fictional participants described below. Keep each person's goals, tone and known information distinct. Stop after six turns and summarize agreements, unresolved issues and assumptions. Do not impersonate a real person or present simulated decisions as approvals from actual stakeholders.

Developer reference

MiniMax API pricing

These are MiniMax API reference prices, not EZ Ai Assist subscription prices.

View EZ Ai Assist plans

Current M2-her API pricing is not verified in the reviewed official pricing tables. Check your provider account or ask MiniMax before estimating cost; rates from MiniMax M2 and other M-series models are not substitutes.

  • An absent price is not a free tier or a zero-cost claim.
  • Confirm billing, legacy endpoint access and applicable limits before moving a dialogue workflow into production.

Common questions

A few things worth knowing.

How is M2-her different from MiniMax M2?

M2-her is the separate dialogue and role-play model. Its ID, role settings, context and dedicated schema differ from the general M2 reasoning model; the names do not imply interchangeable capabilities.

What are its documented limits?

The model guide lists 64K context. The retained dedicated schema gives a 2,048-token maximum completion setting. These references are qualified because current general API documentation does not consistently include M2-her.

Which API endpoint should I investigate?

The dedicated retained schema describes /v1/text/chatcompletion_v2 with model M2-her. Check current account support before using it; the generic Chat Completions documentation now lists other M-series models.

Can I attach tools or enable thinking?

Those fields are not documented in the dedicated schema reviewed here. That is an evidence limit, not proof of every possible deployment's behavior. Do not advertise general M2/M3 tools or controls as M2-her support.

Does legacy mean it has shut down?

No. MiniMax lists it under Legacy Models, but the reviewed sources do not establish a retirement date. Confirm current provider and app access rather than inventing a shutdown schedule.

Why is there no numeric API price?

The reviewed current official pricing tables do not list a verified M2-her rate. Borrowing another model's price would be misleading; confirm the rate in your provider account or with MiniMax.

Can these role settings be used in EZ Ai Assist?

The guide describes provider-level capabilities, not a promise that the app exposes every role field or this legacy endpoint. Check the workspace for available models and controls; ordinary prompt examples do not enable unsupported API features.

Check the source

Official documentation

Specifications and API prices checked on . Example prompts and workflow advice are editorial guidance from EZ Ai Assist.

Same provider

More from MiniMax

MiniMax

MiniMax M3

Multimodal coding model with text, image and video input and up to 1M context, subject to endpoint capacity.

MiniMax

MiniMax M2.7

Text-focused coding and office-work model with 204,800 context, tool calling and always-on thinking.

MiniMax

MiniMax M2.7 Highspeed

Faster-serving M2.7 endpoint with 204,800 context, always-on thinking and separate API rates.

MiniMax

MiniMax M2.5

Legacy coding and tool-workflow model with 204,800 context and always-on thinking.